Error when loading complex NaN's

[DavidWellnitz]

I have a CSV File which contains complex values and some values are NaN + NaN*im, which throws an error when I try to read it with type fixed to ComplexF64. The underlying problem seems to be that parse(ComplexF64, "NaN + NaN*im") throws the error: ERROR: ArgumentError: cannot parse " NaN*" as Float64. I tried to "solve" this by setting strict=false like:
df = CSV.File(fname, header=header, type=ComplexF64, strict=false) |> DataFrame
where fname is the location of my file and header is the header, but it throws the same error.

Am I using strict wrong? And is the behavior of parse intentional?

[DavidWellnitz]

Update: Using missingstring= "NaN+NaN*im" instead of strict=false works, but it still feels like strict=false should also work.

[quinnj]

Hmmm, yes, an interesting problem. For non-standard types, we currently rely on calling Base.parse, but on invalid values, that may throw an error that we're currently not catching. We should probably switch to using Base.tryparse, so that strict=false works as you mentioned. We could also consider adding parsing support in Parsers.jl for Complex types, which would make parsing them much faster.

[quinnj]

Ok, this was fixed in JuliaData/Parsers.jl#86

[DavidWellnitz]

Thanks :)